JP6430290B2 - Magnetic resonance imaging system - Google Patents

Description

  Embodiments described herein relate generally to a magnetic resonance imaging apparatus.

  Conventionally, EPI (Echo Planar Imaging) is known as an imaging method using a magnetic resonance imaging apparatus. Since EPI can acquire a complete 2D image with one shot, it is an effective imaging method when speed is required. Such EPI can cause distortions associated with, for example, magnetic susceptibility, chemical shifts, imperfect magnetic field uniformity adjustments, and eddy currents, which can result in artifacts (eg, ghost artifacts). is there.

US Patent Application Publication No. 13/914160

K.P. Pruessmann et al., “SENSE: Sensitivity Encoding for Fast MRI”, Magnetic Resonance in Medicine, 42, 952-962, 1999 Bruder et al., “Image Reconstruction for Echo Planar Imaging with Nonequidistant k-Space Sampling”, Magnetic Resonance in Medicine, 23, 311-323, 1992 Buonocore et al., “Ghost Artifact Reduction for Echo Planar Imaging Using Image Phase Correction”, Magnetic Resonance in Medicine, 38 (1st edition), 89-100 pages, 1997 Chen et al., "Removal of EPI Nyquist Ghost Artifacts With Two-Dimensional Phase Correction", Magnetic Resonance in Medicine, 51, 1247-1253, 2004 Xu et al., “Robust 2D Phase Correction for Echo Planar Imaging Under Tight Field-of-View”, Magnetic Resonance in Medicine, 64, 1800-1813, 2010 Xiang et al., "Correction for Geometric Distortion and N / 2 Ghosting in EPI by Phase Labeling for Additional Coordinate Encoding (PLACE), Magnetic Resonance in Medicine, 57, 731-741, 2007.

  The problem to be solved by the present invention is to provide a magnetic resonance imaging apparatus capable of providing an image in which artifacts related to differences in image distortion are reduced or eliminated.

  A magnetic resonance imaging (MRI) apparatus according to an embodiment includes a sequence control unit, a generation unit, a correction unit, and a correction unit. The sequence control unit executes a sequence including a pre-scan and a main scan on the subject. The generation unit generates a pre-scan image and a correction map from the data collected by the pre-scan, and generates a main scan image from the data collected by the main scan. The correction unit corrects the correction to have substantially the same distortion as the geometric distortion in the main scan image based on the geometric distortion difference between the main scan image and the pre-scan image. Modify the map. The correction unit corrects the main scan image using the corrected correction map.

FIG. 1 is a high level schematic block diagram of an MRI apparatus suitable for reducing artifacts associated with EPI according to one embodiment. FIG. 2 is a flowchart of a technique for reducing artifacts in EPI according to an embodiment. FIG. 3 is a schematic diagram illustrating estimation of a two-dimensional (2D) distortion map according to one embodiment. FIG. 4 is a diagram illustrating a process of expanding FOV (Field Of View) in a pre-scan image according to an embodiment. FIG. 5 is a graph showing the estimation of the shift amount for correcting the main scan image according to the embodiment. FIG. 6 is a graph showing the estimation of the shift amount for correcting the main scan image according to the embodiment. FIG. 7 is a diagram illustrating examples of a ghosted image, a conventional ghost correction map, and a distorted ghost correction map according to an embodiment.

  Hereinafter, an embodiment of an MRI apparatus will be described in detail based on the drawings.

The MRI apparatus 20 shown in FIG. 1 includes a gantry 10 (shown in a schematic cross section) and various related system components connected thereto. At least the gantry 10 is usually arranged in a shield room. The structure of the MRI apparatus 20 shown in FIG. 1 includes a static magnetic field (B ₀ ) magnet 12 arranged in a substantially coaxial cylindrical shape, a gradient magnetic field coil 14 of Gx, Gy, and Gz, and a large RF coil. It has a whole body coil (WBC) 16. An imaging volume 18 is shown so as to substantially surround the head of the subject 9 supported by the subject (patient or the like) bed 11 along the horizontal axis of the cylindrically arranged elements. An array coil (AC) 19, which is one or more small RF coils, is placed on the head of the subject (hereinafter referred to as “scan target” or “subject”) in the imaging volume 18. You may combine in close proximity. As will be apparent to those skilled in the art, a coil or AC that is small compared to a WBC, such as a surface coil, can be used for certain body parts (eg, arms, shoulders, elbows, wrists, knees, legs, chest, spine, etc. ) Is often designed to match. Hereinafter, such a small RF coil is referred to as an AC or a phased array coil (PAC). These are configured to receive at least one coil configured to transmit an RF signal into the imaging volume 18 and an RF signal from a subject such as the head of the subject in the above example in the imaging volume. And a plurality of receiving coils.

  The MRI system control unit 22 has an input / output port connected to a display 24, a keyboard 26, and a printer 28. As a matter of course, the display 24 may be of a touch screen type so that control input can be performed, or an input / output device such as a mouse may be provided.

  The MRI sequence control unit 30 is connected to the MRI system control unit 22, and includes Gx, Gy, and Gz gradient coil driver 32, and an RF transmitter 34 and a transmission / reception switch 36 (the same RF coil is used for both transmission and reception). Control). The MRI sequence controller 30 includes a suitable program code structure 38 for implementing MRI imaging (also known as Nuclear Magnetic Resonance (NMR) imaging) techniques including parallel imaging. The MRI sequence control unit 30 may be configured for EPI or parallel imaging. Further, the MRI sequence control unit 30 may perform one or more preparation scan (pre-scan) sequences and a scan sequence for acquiring a main scan image (sometimes referred to as a diagnostic image). For example, a sensitivity map (sometimes referred to as a coil sensitivity map or a spatial sensitivity map) for the RF coil 16 or the RF coil 19 may be generated using the MR data from the pre-scan, and parallel imaging is developed. A return map may be generated. In one embodiment, a prescan, referred to herein as an EPI prescan, is performed for one or more EPI main scans.

  The MRI apparatus 20 has an RF receiver 40 that sends input to the MRI data processing unit 42 in order to create processed image data to be sent to the display 24. In addition, the MRI data processing unit 42 may generate MR data and images generated before, for example, a coil sensitivity map, a folding map for developing parallel imaging, a ghost reduction map, and a distortion map (for example, according to an embodiment). Configured to access a map such as a map showing the difference between a pre-scan image and a main-scan image), system configuration parameters 46, and MR image reconstruction program code structures 44 and 50.

  FIG. 1 shows a general description of the program storage device 50 of the MRI apparatus. In program storage 50, stored program code structures (eg, for image reconstruction with reduced or eliminated ghost artifacts, or for defining a graphical user interface and receiving operator input thereto) are stored in MRI. Stored in a non-transitory computer readable storage medium accessible to the various data processing components of the device. As will be apparent to those skilled in the art, the program storage 50 is segmented to require at least a portion of such stored program code structure in normal operation among the processing computers of the MRI apparatus 20 with highest priority. It may be connected directly to another computer (ie, not normally stored or directly connected to the MRI system controller 22).

  Indeed, as will be apparent to those skilled in the art, FIG. 1 shows a very high level schematic of a typical MRI apparatus that has been modified to implement the exemplary embodiments described below. The components of the system can be divided into “boxes” of various logical sets, usually with a large number of digital signal processors (DSPs), microprocessors and dedicated processing circuits (eg for high-speed AD conversion). , For fast Fourier transform, for array processing, etc.). Each of these processors is typically a clocked “state machine”, and the physical data processing circuit moves from one physical state to another every clock cycle (or a predetermined number of clock cycles). .

  Not only does the physical state of a processing circuit (eg, CPU, register, buffer, computing device) gradually change from one clock cycle to another, but also the physical state (eg, magnetic) of the associated data storage medium. Bit storage locations in the storage medium are also converted from one state to another during the operation of such a system. For example, at the end of the image reconstruction process and sometimes the image reconstruction map (e.g., coil sensitivity map, unfolding fold map, ghost map, distortion map) generation process, computer readable and accessible in physical storage media The array of data value locations is converted from a certain previous state (eg, all uniformly “0” values or all “1” values) to a new state. The physical state of the physical location of such an array varies between minimum and maximum values, real-world physical events and physical conditions (eg, the internal physical structure of the subject in the imaging volume space). Represents. As will be apparent to those skilled in the art, a specific structure of computer control that causes a particular sequence of operating states to cause and transition within the MRI apparatus when sequentially read into the instruction register and executed by one or more CPUs of the MRI apparatus 20 Similar to program code, such an array of stored data values represents and constitutes a physical structure.

  The exemplary embodiments described below provide EPI images with reduced or eliminated artifacts related to image distortion differences. Although many of the embodiments described herein are for EPI, the techniques described herein for removing or reducing artifacts related to image distortion differences are MRs other than MR images collected by EPI. It may be applied to images (MR images acquired by other MRI pulse sequences).

  An MR image is formed by collecting a nuclear magnetic resonance (NMR) RF response signal (eg, echo data) that is spatially encoded for each corresponding point in k-space. The RF response signal is typically generated by “traversing” k-space in two or three dimensions according to a set MRI pulse sequence. Acquisition of echo data in the frequency encoding spatial encoding direction (eg, along the x-axis) is usually fast and is on the order of a few milliseconds. However, each point is sampled with a different value of the applied phase encode field gradient along the phase encode axis (eg, the y axis). Therefore, generally, the acquisition time of the MR image is mainly determined by the number of phase encoding steps.

  In parallel imaging, the acquisition time can be shortened by undersampling the k-space along the phase encoding direction. In many parallel imaging techniques, the acquisition time is significantly reduced because the number of k-space points sampled along the phase encoding direction is reduced. Spatial RF transmit / receive patterns that provide spatially encoded signal strength and coil positioning information specific to the RF coil (eg, PAC) by parallel imaging techniques instead of sampling each point along the phase encoding direction And the MR image is reconstructed with fewer samples of selected points along the phase encoding direction. Also, spatial resolution can be improved in images reconstructed based on parallel imaging. Non-Patent Document 1 describes a parallel imaging and reconstruction technique, the contents of which are incorporated herein in their entirety by reference. Another example of parallel imaging and reconstruction technology is SPEEDER® available on Toshiba MRI machines.

  EPI is another imaging technique to reduce acquisition time. EPI is often used in MRI where speed is critical. The application of EPI has the advantage that a complete 2D image can be acquired with a single shot. When a 2D k-space encoding is collected from a single excitation or one spin interference path, a 2D acquisition result (or 3D acquisition result) is generated. Collecting a complete 2D image in a single shot, for example, does not require compensation for subject movement (eg, subject movement between collection shots) or reduces the need for compensation There is an effect that is done. In EPI, generally, one row of k-space encoded data is collected at a time in about 1 millisecond, and 2D array k-space encoded data is collected at a time in about 20 to 120 milliseconds.

  Acquiring a complete image in a single shot includes, for example, diffusion (Diffusion Weighted Imaging (DWI), etc.), functional MRI (Functional MRI: FMRI), and perfusion (dynamic susceptibility contrast ( The application of Dynamic Susceptibility Contrast (DSC) augmentation or Arterial Spin Labeling (ASL) is highly desirable, and MR manufacturers often require the use of diffusion-weighted EPI for possible acute cerebral hemorrhage (stroke) ) For diagnosis and the like.

  However, EPI has drawbacks such as susceptibility, chemical shift, imperfect magnetic field uniformity adjustment, and large distortion associated with eddy currents. Gradient magnetic field eddy currents may be generated by gradient magnetic field hardware design, RF coils, shields, etc., which causes ghosts in EPI. Inadequacies such as an asymmetrical RF reception frequency response cause spatially dependent EPI ghosts. Conventional acquisition encoding and reconstruction can cause such a spatial misalignment of distortion.

  Spatial misregistration can occur with any type of MRI, but is often particularly noticeable in EPI. For efficiency reasons, a plurality of different lines in k-space are sampled using bi-directional EPI main scan readout. Encoding mismatch between two alternating polarity readouts introduces errors in standard reconstructions, and in particular causes a typical “Nyquist ghost”, also known as “N / 2 ghost”.

  In particular, when the k-space is traversed in a high-speed zigzag pattern (for example, the readout (RO) gradient magnetic field is alternately replaced and crosses the k-space back and forth), the y-direction (for example, a normal phase encoding blip is used) Phase encoding direction) is traversed at a slower rate than the x direction (eg, frequency encoding direction). This may cause the second gradient magnetic field (eg, in the phase encode direction) to be significantly weaker than the first gradient magnetic field, resulting in greater distortion along the phase encode direction. If the signal phase is unintentionally accumulated at a low speed, abnormal phase encoding, in particular, spatial positional deviation in the phase encoding direction of the image occurs.

  Parallel imaging is often included in EPI to reduce the inevitably occurring distortion and maintain spatial resolution, especially at higher magnetic field strengths (eg, 3T). The parallel imaging method can reduce the influence of unintended signal phase accumulation on intentionally applied phase encoding. For this reason, geometric distortion can be reduced to about 1/2 or 1/3 in many cases with the acquired parallel imaging acceleration factor.

  However, in parallel imaging, other distortion artifacts such as reconstruction artifacts may occur in the output diagnostic image. The reconstruction artifacts are mainly due to the development process required to obtain the desired diagnostic image based on the intermediate image obtained by undersampling (to reduce the image reconstruction area) in the phase encoding direction. . “Development” is a process of combining a plurality of folded images in order to generate a desired diagnostic image. In many cases, the multiple folded images differ from each other in that each is collected from a different RF receiver coil.

  Many parallel imaging techniques reconstruct the final diagnostic image using two different image acquisition methods. In addition to the main scan, the coil calibration prescan image is also useful for the final diagnostic image. In the main scan, since the image reconstruction area that has not been folded is narrow, the pixels are folded. Depending on the main scan, for example, in the EIP main scan, the MR signal is displaced (or shifted) before the aliasing prevention is performed as compared with the corresponding coil calibration scan image (or the corresponding coil sensitivity map). )Sometimes. If this defect is not solved, the development artifact may appear in the final diagnostic image. The most serious of these artifacts appears as a region of discontinuous aliasing errors shifted in the phase encoding direction. These artifacts can also be due to imperfect signals or increased noise (reduced signal-to-noise ratio (SNR)) at certain locations, or some or all of such features. May appear as a combination.

  In the main scan image, it is often desirable that the ghost artifact level be low, and in some cases less than 2% of the intensity of the “right” or “ghost-free” signal. Therefore, effective 2D spatial dependence correction is required when there is susceptibility distortion or off-resonance inherent EPI distortion.

  Ghost artifacts in EPI images can be corrected to some extent with various technical costs such as lengthening the acquisition time or eliminating time encoding by single shots. For example, a conventional technique such as reducing a ghost using a scalar estimated value of a data correction coefficient may be used. In many cases, zero order constants and first order constants can be applied to improve the consistency of the two readout polarities in the image.

  Another conventional technique for correcting ghost artifacts in EPI, or in parallel imaging, or in EPI combined with parallel imaging, is to apply spatial variation to more advanced ghost correction. 1D correction can be estimated and applied using techniques described in Non-Patent Document 2 and Non-Patent Document 3. Further, 2D correction can be estimated and applied using the techniques described in Non-Patent Document 4, Non-Patent Document 5, and Patent Document 1. Patent Document 1 is incorporated herein in its entirety by reference.

  Inverting EPI distortion using time-shifted multiple acquisitions or multi-directional phase encoding is an example of a technique for correcting EPI geometric distortion. Non-Patent Document 6 describes an example of a technique using a plurality of time-shifted acquisitions. If the EPI image is distorted, the signal may “stack”, multiple locations may be mapped to one distorted location, or a large area of the signal may be compressed into a much smaller area There is also a possibility that. “Inverting” such an area may not be possible with a single EPI shot.

  In some embodiments, parallel imaging correction is performed using a 2D parameter map of coil sensitivity, and Nyquist ghost correction is performed using a 2D map of ghost correction values having phase correction parameters or complex values. These ghost correction maps are based on MR data collected by prescan.

  If the ghost correction map associated with the EPI main scan exhibits a geometric distortion that differs from the geometric distortion of the main scan image, an actual spatial misalignment exists between the data sets in the reconstruction.

  Embodiments described herein estimate a distortion difference between one or more pre-scan images and a main scan image, so that the pre-scan image has substantially the same amount of distortion as the main scan image. Deliberately distort the scanned image. Finally, reconstruction is performed based on one or more intentionally distorted pre-scan images and main scan images.

  In one embodiment, the same EPI prescan can be used to determine the ghost correction map and the intentionally distorted prescan image. Furthermore, the EPI pre-scan may be a pre-scan that is normally executed, and it is not necessary to increase the total acquisition time (that is, the throughput of the subject). Further advantages of the embodiments are given below, including the end of this description before the claims.

  FIG. 2 shows a flowchart of a method 200 for reducing or eliminating ghost artifacts in an MRI apparatus using EPI according to one embodiment. Of course, one or more of the steps 202-228 may not be performed and may be performed in a different order than that shown in FIG. For example, the method 200 may be executed and controlled by one or more of the MRI data processing unit 42 and the MRI system control unit 22, such as the program storage device 50, the MRI image reconstruction program code structure 44, the image memory 46, etc. Memory may be accessed.

  After entering a subroutine corresponding to method 200, step 202 is performed by the MRI apparatus to collect MR data sets corresponding to at least one EPI main scan and one or more prescans including an EPI prescan. Executed.

  Step 202 includes collecting EPI main scan (often accelerated by parallel imaging) MR data. An EPI main scan includes a significant amount of geometric distortion, as described above. EPI main scans usually contain ghost artifacts when reconstructed according to the prior art. Such ghost artifacts are desirably reduced or eliminated before the final MR image used for diagnostic purposes or the like is generated.

  Step 202 also includes collecting EPI pre-scan data sets and parallel imaging coil sensitivity pre-scan data sets (often performed prior to EPI main scan). These pre-scans usually have a different amount of distortion than the EPI main scan.

  The EPI pre-scan is performed with a low phase encoding intensity so that the imaged subject contracts in the FOV and consequently does not overlap in the FOV due to either ghost artifacts or parallel imaging (this is , FOV may be regarded as spreading in the phase encoding direction). In fact, this effect can be obtained by reducing the EPI pre-scan blip (for example, phase encoding gradient magnetic field) to about 1/2 to 1/3 of the EPI main scan blip. Since EPI prescan is performed without parallel imaging, the EPI prescan blip strength is the parallelization factor used in parallel imaging (eg, 1/2 to 1/3 in many applications). It may be further reduced. Therefore, depending on the embodiment, the EPI pre-scan blip may be reduced to, for example, ¼ to 1/9 with respect to the EPI main scan including parallel imaging. By clearly separating the object to be imaged from the ghost in the EPI pre-scan, the amount of signal in the object and the amount of signal in the ghost can be determined, and the signal of the main scan image and the ghost image at various positions Further useful information such as a relative phase difference from the signal of the signal can be obtained.

  In step 202, MR data is also obtained. In step 204, an unfolded image of the EPI main scan is generated from the MR data. The EPI pre-scan MR data obtained from step 202 is used to generate one or more coil sensitivity maps in step 206, a folding map for development in step 208, and a pre-scan image in step 210. . The coil sensitivity map may be used when developing from parallel imaging. Conventional techniques may also be used in step 202, which is conventional for forming the main scan image in step 204, the coil sensitivity map in step 206, the folding map in step 208, and the prescan image in step 210. Techniques may be used.

  According to one embodiment, as an input to a new refinement process to correct for geometric distortion differences by techniques to remove or reduce artifacts, particularly artifacts due at least in part to different geometric distortions. A folded main scan image is obtained from step 204, and one or more images are obtained from steps 206, 208, and 210. An example process for correcting for geometric distortion differences involves pixel shifts as needed to align two of the images from steps 204-210 (eg, corresponding to differently distorted images, respectively). The amount of pixel shift) is calculated. Preferably, the shift amount is determined based on the main scan image and the pre-scan image having different distortions. A relative distortion map is obtained from the shift amount for each pixel determined in this way.

  The EPI main scan and the EPI pre-scan have different phase encoding gradient magnetic field strengths. The amount of distortion, that is, the distance by which the elements of the main scan image and the prescan image are shifted in the phase encoding direction is related to the strength of the gradient magnetic field. For example, consider an EPI pre-scan and an EPI main scan in which the main scan phase encode blip is six times the pre-scan phase encode blip. In this example, the distortion in the pre-scan image can be six times larger in the phase encoding direction than the distortion in the EPI main scan. The distortion in the phase encoding direction is a geometric distortion that is dominant. The background frequency error has a larger influence on the phase encoding gradient magnetic field as the phase encoding gradient magnetic field is weaker.

  In step 212, a ghost correction map is generated from the data collected by the EPI prescan. The ghost correction map is basically a smooth and slowly changing function. Depending on the embodiment, the ghost correction map may be generated by using the technique described in Patent Document 1 incorporated by reference.

  In step 214, an intermediate developed image is generated using one or more of the coil sensitivity map, the unfolding return map, and the ghost correction map generated in steps 206, 208, and 212, respectively. The first expansion of the main scan image may be executed according to the technique described in Patent Document 1 incorporated by reference.

  In step 216, a pre-scan image having a low resolution in the phase encoding direction is generated. Conventional techniques may be used to determine the prescan image. These techniques may include geometrically stretching a pre-scan image collected with a smaller phase encoding gradient. These techniques may also include separating the main image from non-overlapping ghost images, and appropriately recombining the ghost with the main image.

  The intermediate development image generated in step 214 and the pre-scan image generated in step 216 are used when generating the first distortion map in step 218.

  As described above, the distortion map is obtained by quantifying the amount of distortion of a predetermined reconstructed image for each pixel so that the geometric distortion substantially corresponds to the geometric distortion of the intermediate development image. is there. In one embodiment, a 2D distortion map may be used that represents that a selected EPI pre-scan shot (ie, template) is shifted from the EPI main scan in the phase encoding direction. FIG. 3, described below, illustrates shift estimation in a distortion map according to one embodiment. FIG. 4 described below shows a step of expanding the FOV of the pre-scan image so that it can be used in distortion estimation according to an embodiment.

  In one embodiment, an image with moderate distortion (e.g., equivalent to MR data collected by a stronger phase encoding gradient in an EPI main scan) and an image with greater distortion (e.g., EPI pre-scan). In the two images (corresponding to the MR data collected by the weaker phase encoding gradient field), the amount of distortion difference may be determined by identifying corresponding points. The corresponding points in the two images can be identified using line integration in some embodiments. FIGS. 5 and 6 described below show a process for determining a difference in distortion at a certain point, that is, a necessary amount of shift, using a line integration method.

  In addition, the ghost correction map may be distorted using the first distortion map. In step 220, the first distortion map is changed in magnification so that the coil sensitivity map and the unfolding map for development are estimated to be distorted with respect to the main scan image. Two distortion maps may be generated.

  In order to make such a magnification change, it may be beneficial to first associate the sensitivity to frequency-induced geometric errors with each of the image acquisitions by step 202. For example, when the acquisition by the EPI main scan is performed with a predetermined average phase encoding gradient magnetic field strength, this strength may be arbitrarily designated as a sensitivity of “1” to the geometric distortion. Then, an EPI pre-scan with the phase encoding gradient reduced to 1/6 would have a relative geometric distortion of “6” from the shift with frequency. The coil sensitivity map that would not be collected by EPI may have zero effective distortion in the phase encoding direction due to the frequency shift mechanism. Therefore, the distortion difference between the EPI main scan and the EPI pre-scan is “5” (that is, 6 minus 1 in this example). The distortion difference from the EPI main scan to the coil sensitivity map is −1 (ie, 0 minus 1).

  In this example, the ratio is set to -1 / 5 = -0.2 in order to change the required amount of strain induced accordingly. Therefore, the pixel value of the second distortion map is obtained by multiplying the distortion value of each pixel in the first distortion map by −0.2. A possible improvement in the embodiment can be considered by arbitrarily taking into account other distortion mechanisms such as the accompanying gradient field effect, eddy currents due to large diffusion-encoded gradient fields, and slow frequency shifts due to heat. It is done.

  In step 222, the coil sensitivity map generated in step 206 is intentionally distorted using the second distortion map. Thereby, the distorted coil sensitivity map has a distortion substantially equal to the distortion of the main scan image.

  In step 224, the return map generated in step 208 is intentionally distorted using the second distortion map. Thereby, the distorted return map has a distortion substantially equal to the distortion of the main scan image.

  In step 226, the ghost correction map generated in step 212 is intentionally distorted using the first distortion map. Thereby, the distorted ghost correction map has a distortion substantially equal to the distortion of the main scan image.

  FIG. 7 described below shows an image with a ghost as generated based on MR data collected by an EPI main scan of a subject located in an imaging volume, and one or more pre-scans of the subject. FIG. 5 shows a conventional ghost correction map based on and an intentionally distorted ghost correction map according to one embodiment.

  In step 228, the folded main scan image obtained from step 204, (a) the intentionally distorted coil sensitivity map generated in steps 222, 224 and 226, respectively, (b) the intentionally distorted folded map. And (c) one or more of the intentionally distorted ghost correction maps are used to reconstruct the final undeveloped ghost-reduced EPI diagnostic image. This is the second development of the main scan image. The first expansion is performed without geometric distortion correction, whereas the second expansion is performed with geometric distortion correction. The second expansion includes ghost removal using an intentionally distorted ghost correction map, and expansion / reconstruction using an intentionally distorted coil sensitivity map and an intentionally distorted aliasing map, where All of these maps have a geometry that matches the EPI main scan.

  Note that in the described embodiment, attention is directed to the relative geometric distortion between the components of each image or map during the reconstruction process. In this case, it is not necessary to geometrically correct any distortion that occurs in the main scan image itself. Embodiments herein can be used in combination with a process that attempts a clear geometric correction of a distorted main scan image, or can be used alone without any combination. Such distortion correction of the main scan image has been described in the literature and involves problems and advantages that are largely unrelated to the methods described herein. The embodiments described herein are primarily based on discrete effects of the EPI signal, for positions related to the effects of altering the readout gradient (Nyquist ghost), parallel imaging artifacts, or a combination of these two artifact mechanisms. It relates to reducing aliasing.

  FIG. 3 shows a schematic diagram illustrating an estimation of a 2D distortion map, such as the first distortion map or the second distortion map described above, according to one embodiment. As described above, in one embodiment, the distortion map indicates how much the selected EPI pre-scan shot (also referred to herein as a “template”) is shifted from the EPI main scan in the phase encoding direction. Show.

  A schematic diagram 302 shows a pre-scan image obtained from an EPI pre-scan shot. A schematic diagram 304 correspondingly shows a schematic diagram of a main scan image obtained from the EPI main scan. Usually, the EPI main scan is performed after the EPI pre-scan. The main scan image may be formed after the development process. Diagram 306 shows the estimated 2D distortion map. A 2D distortion map 306 may be obtained by comparing the developed main scan image 304 with the pre-scan image 302 to determine the distortion difference at each location.

  The direction of the arrow in the schematic diagram 306 indicates the direction of distortion at each point (eg, pixel), and the length of the arrow indicates the magnitude of distortion (eg, the magnitude of shift). Thus, the distortion template or distortion map includes the vector amount (direction and magnitude of the shift value) at each position. The diagram 306 shows, for example, that the EPI pre-scan is shifted by various amounts in the phase encoding direction compared to the main scan image. Assuming that the shift direction is constant, only the shift amount data needs to be shared in the shift template.

  FIG. 4 shows a process of expanding the FOV in the pre-scan image according to one embodiment.

  As shown in diagram 402, the original pre-scan image is “crushed”. The squashing of the image is due to the small phase encoding blip used in the EPI pre-scan, so that the actual image of the subject is a Nyquist ghost as shown here, as shown schematically in FIG. Clearly separated from ghosts etc.

  Since the pre-scan image 402 can be expanded by a set or dynamically determined expansion coefficient so as to have the same FOV as the main scan image, it can be used for distortion estimation. Stretching by a factor of 2 to 4 is common, but other factors are possible and contemplated in embodiments. Diagram 404 shows prescan image 402 after it has been stretched to full FOV.

  When multiple receive coils are used, in practice it is usually necessary to combine the images from the individual coils in order to create a ghost correction map suitable for use throughout the image. When a magnitude image is required and a combination of parallel imaging such as SENSE is not used, it is easy to combine images of each coil by “sum of squares”, which is very effective. However, different parallel imaging reconstructions include different effective coil combination mechanisms, ie different spatial intensity weighting implications. Therefore, when multiple images are used in the calculation, it may be beneficial to consider whether different spatial combination techniques, i.e., different spatial weights, can cause problematic intensity differences.

  In one example, the pre-scan image from step 216 may be weighted with “Sum Of Squares”.

Furthermore, the development process in step 214 may use a relative sensitivity map when a coil ratio is formed between the received image of the array coil σ _{aci and} the received image of the relatively uniform WBC.

  As is well known, the overall spatial weighting of such a developed image is substantially that of the WBC image.

    SENSE development uniformity = WBC (3)

  And in order to perform the operations between these two image results on a pixel-by-pixel basis, one or both of the images may need to be modified so that these images share a single spatial weight. In this example, such correction for one correction applies a spatial enhancement correction to the developed image from step 214 for each pixel as follows.

  Of course, the exact correction in various coil combination techniques, i.e. the correct correction in various parallel imaging algorithms, depends entirely on the specific combination method and the specific parallel imaging method.

  FIGS. 5 and 6 illustrate techniques for determining shift amounts using line integration methods, according to some embodiments. With line integration, the total amount of signal along the vertical line is preserved so that if the image is partially stretched, the signal along the vertical line may shift up or down the line. For this reason, in a selected pixel of an image with little distortion, for example, 45% of the signal is below the point selected during line integration, and 55% of the signal is above this point. . In a normalized line integral curve corresponding to an image with low distortion, the amplitude of such a point is 0.45. A point in the normalized line integral of an image with higher distortion that has the same percentage of signal as an image with moderate distortion (for example, 45% is less than this point and 55% is higher than this point in this example). By identifying, it is possible to identify a corresponding pixel in an image with large distortion. And the amount of shift between the two images is the difference between the positions of the two corresponding pixels. In the method described above, points corresponding to all signal values can be determined.

  This may be repeated for each pixel. In some embodiments, the above process may be repeated only for every nth pixel in a row or column, where n is 2 or greater. The above technique is not limited to estimating the shift distance only from the center position of the distant pixels. An intermediate position between distant pixels may be determined by interpolation. Similarly, the techniques described above are not limited to determining shift displacements that are individual multiples of the pixel size. In practice, the amplitude values of two distant pixels are a combination of the amplitudes of the displacements of the two distant pixels at a certain rate.

  Interpolating to the level of the pixel portion helps to avoid stepping effects that can occur in the corrected image, given that the correction is limited to applying all the displaced pixel displacements.

  According to one embodiment, the distortion map is formed by repeating the above for each pixel. For each position, the distortion map includes a positive or negative shift corresponding to the number of pixels in the phase encoding direction.

  The schematic diagrams 502 and 504 are graphs showing signal values for each column, for example, for each row in the phase encoding direction, in each of the uniform pre-scan image and the intermediate developed image of the main scan image. Graphs 502 and 504 are identified as “T” (eg, indicating a template) and “M” (eg, indicating an EPI main scan), respectively.

  Schematic diagrams 506 and 508 show the result of integrating “T” and the result of integrating “M”, respectively, along the phase encoding direction. When the integration of T shown in the graph 506 and the integration of M shown in the graph 508 are respectively determined, the shift at each position can be estimated. For example, the amplitude a is obtained for each position y on the curve representing the integral of M. A point corresponding to the amplitude a is identified in the curve representing the integral of T. Next, the position yz is identified in the curve representing the integral of T. Then, the shift in y is determined as (y-yz).

  FIG. 6 shows a more detailed view of the estimation outlined in FIG. 5 according to one embodiment.

  A schematic diagram 602 shows a curve 606 representing the integral of T and a curve 604 representing the integral of M in the same graph. First, the value of the curve 604 corresponding to the position y on the phase encode axis is selected as the amplitude a. As a next step, a position yz on the phase encode axis corresponding to the amplitude a on the curve 606 is identified.

  Schematic diagram 612 shows a magnified view of a portion of the areas of curves 604 and 606 that are most relevant to determining yz. Starting from the point 614 corresponding to y, a point 616 representing the amplitude a on the curve 604 is determined. Of course, this point is the position corresponding to the phase encode value of y on the curve 604. Next, a point 618 on curve 606 having the same amplitude as point 616 is identified. On the graph shown, point 618 may be identified by drawing a horizontal line from curve 604 graphically, which intersects curve 606 at point 618.

  After the point 618 is identified, the corresponding position yz on the phase encode axis is determined. Point 620 represents yz corresponding to point 618 on curve 604.

  Therefore, the shift at point 614 is the difference between point 614 and point 620. In other words, the shift at point y can be expressed as y-yz.

  FIG. 7 shows an image 702 with a ghost. The ghosted image 702 is generated based on MR data collected by an EPI main scan from a subject located in the imaging volume, for example. FIG. 7 shows a conventional ghost correction map 704 generated based on one or more pre-scans of the subject according to the conventional technique.

  FIG. 7 shows a second ghost correction map 706. The second ghost correction map 706 is formed by modifying the conventional ghost correction map 704 to have the same (or nearly the same) distortion as the ghosted image 702, according to one embodiment. It can be seen that the similarity between the ghosted image 702 and the modified ghost correction map 706 is higher than the similarity between the ghosted image 702 and the conventional ghost correction map 704. By modifying the ghost correction map, the ghost correction map has a geometric shape that matches the EPI main scan.

  In the described embodiment, the readout direction is not corrected. The distortion amount in the readout direction is extremely small compared to the distortion in the phase encoding direction and can be ignored. However, the embodiment is not limited to the case where the distortion in the phase encoding direction is corrected.

  In the above-described embodiment, the originally smooth ghost correction map is shifted and distorted. This arrangement has the advantage that errors due to relatively small amounts of noise in the data do not cause significant errors in the final image. For example, using a smooth ghost correction map to cause a 2.8 pixel shift instead of 2.6 pixels as actually required, the ghost correction map is off by 0.2. However, since it is a smooth map, this does not cause significant errors in the image. Only very small, undetectable errors in the second order or final image occur. On the other hand, when a main scan image that can have sharp image characteristics (for example, an image feature amount that changes rapidly between two adjacent pixels) is shifted and distorted, the error described above is an error of 20%. appear. Due to the error of 20%, a sharp boundary changes suddenly, and a ghost may occur.

  The prior art attempts to remove distortion of the main scan image itself by, for example, imaging two main scan images with different phase encodings. However, in such a technique, an error caused by scanning at different moments may occur, and the superiority of the single shot EPI is lost to some extent.

  If there is 10% ghost before, during and after effective Nyquist correction, there will be 1% error. Here, it is assumed that a slight error occurs in the Nyquist map, and a shift of 2.6 pixels occurs instead of a shift of 2.2 pixels. This may result in a ghost that can be an error of 1.1% or a ghost error of 2%, but rather than projecting an error of about 40% of a clear boundary on a diagnostic image (main scan image). Much better.

  In an embodiment, an algorithm for reconstructing a main scan image with reduced or eliminated artifacts is used, for example on 2D slices. Each slice may be reconfigured individually. The “slice” to which the reconstruction algorithm is applied may be spatially different geometric slice locations (as is often referred to as 2D multi-slice imaging), and is common in contrast-enhanced diffusion imaging Or different time iterations in the dynamic imaging process (as is common in BOLD functional imaging, brain activity, etc.). Alternatively, different slices may be applied to repeatedly image the same geometric position. However, it is applied with different signal contrast mechanisms added before EPI readout, as is common in diffusion-weighted imaging, or diffusion tensor imaging, or diffusion fiber tracking applications.

  When the reconstruction algorithm is applied to different types of slices (such as time series or different DWI signal weights), EPI pre-scan shot collection is performed only once for each EPI main scan shot at a predetermined geometric position. And may be useful. Therefore, several 2D images (generally referred to as “slices” in this specification) may be used as a set for subsequent processing and analysis. By taking advantage of a single, or common EPI pre-scan collection, for each “slice” reconstruction (at a given physical location), at least two advantages are obtained.

  The first advantage is that the time required for pre-scanning is small. The second advantage is that using the same EPI pre-scan for EIP main scans at all times in the time series, no additional changes are imposed on the reconstruction process, and in the time series of EPI main scan reconstructions. The variation can be interpreted as due only to the EPI main scan in the time series of acquisition, as opposed to due to a somewhat vague combination of variations in both EPI main scan and pre-scan (s).

  In an embodiment, prescan is used with one or more EPI main scans without sacrificing the EPI feature that one output image is obtained in one acquisition. Therefore, considering the time series effect, it is advantageous over the conventional technique that sacrifices the feature that one output image can be obtained by one acquisition in enabling high-speed scanning of a series of DWI contrasts. .

  The data for each physical coil is handled individually. For each coil, the collected k-space data is divided into two data sets, one corresponding to the positive readout gradient and the other corresponding to the negative readout gradient.

  The k-space data for one read polarity is “inverted” during reconfiguration with respect to the other read polarity. This is well known in EPI. For example, if even rows are considered to be collected with a “positive” readout, perhaps the odd rows are inverted before applying a Fourier Transform (FT) reconstruction operation to the data. Alternatively, the odd image may be inverted in the reading direction after 2DFT or the like.

  A convenient way to divide the data into even k-space sets and odd k-space sets is to copy the entire data set to have two different data sets. For what would be an even set, one could start with the entire data set and replace all of the odd rows with (complex) zero values. To generate an odd data set, starting from the entire data set, all even row data may be replaced with zeros. Other techniques for generating odd and even data sets are possible and are contemplated within the scope of the embodiments.

  EPI pre-scan data is collected with a shorter phase encoding blip, especially compared to the conventional amplitude of blips used in a general FOV EPI scan. Shorter blips are probably shorter by a factor in the range of 2.0-3.0, and the apparent size of the subject decreases in the phase encoding direction by a corresponding factor of 2.0-3.0. . The purpose is to shrink the main scan image and sufficiently shrink the Nyquist ghost image so that these images do not overlap in the reconstruction. Thus, with normal reconstruction (before or without splitting into even and odd parts), multiple images with one central portion of the main scan image and the outermost image of the Nyquist ghost signal are obtained. . By separating the main scan image from the ghost (preventing overlapping pixels), the two parts can be formed in a clear ratio. An example of this operation is disclosed in Non-Patent Document 5 above.

  Processing steps are then used to (A) calculate the ratio of each pixel of the main scan signal to the Nyquist ghost signal, (B) limit this ratio to represent complex phase changes rather than complex amplitude information, and (C) Information from individual RX coils may be combined. However, the embodiment can be considered by excluding any of (A), (B), and (C). For example, Patent Document 2 describes an advantage of excluding a step of limiting a ghost correction map to only phase information, an advantage of maintaining an individual ghost correction map for each coil, and the like. Incorporated herein.

  In the embodiment, a function (A) for determining a phase ratio for each pixel (for example, when the complex complex number is multiplied, the same complex phase angle as the complex number division is generated), and the coil data is set to a specific amount of amplitude weighting in the combination step. Both combined functions (C) are executed. Moreover, the function (B) which restrict | limits the ratio with respect to phase angle information is performed instead of size operation.

  In addition, the ghost correction map is stretched. Prescan phase encoding blips are reduced in intensity, but nominally arrange a k-space array in a matrix of a predetermined size and spacing compared to the conventional method of collecting scans of the entire image reconstruction area. That has not changed. Therefore, phase encode blips are intentionally “inappropriate” because they are collected with less gradient encoding, possibly with a factor in the range of 2-3 than the conventional full imaging area, so that the conventional 2DFT reconstruction An image in which the nominal subject size is compressed with respect to the user-defined phase encoding FOV is obtained.

  In order to take advantage of the geometrically compressed ghost correction map “PH”, the complex-valued ghost correction map “PH” is resampled and with the same coefficients that the prescan blips are shortened, eg It is stretched in the phase encoding direction, possibly by a factor in the range of 2-3.

  The nominally appropriate FOV ghost correction map may include a spatially varying distortion different from the spatially varying distortion in the main scan image or the like.

For each receiving coil (RX coil), two ideal “pseudocoils” are generated. The pseudocoil is determined such that the difference rate of the complex ratio of the two pseudocoils corresponds to the difference between the even and odd images (ie, positive and negative readouts).
In this example, an even number of pseudo-coil sensitivity maps (subscripted here as 2 * coil-1) are collected in SENSE prescan or SPEDER prescan (not to be confused with EPI prescan). Corresponds to the sensitivity map. In this example, an odd number of pseudocoils (here, subscripted 2 * n) are formed by applying an additional ratio factor as calculated in line 8 of the pseudocode. .

  The input sensitivity coil map has a nominally appropriate total FOV but is different from the spatially varying distortion in the main scan image and may be different from the distortion in the EPI pre-scan image. It may include distortion.

  The nominally compressed prescan image is stretched and converted to a composite image or a composite image. Here, the composite image is a composite of the sum of squares. Other types of coil combinations may be considered, especially if the same effective final composite weighting is later applied in the main scan image. For example, these composite images may or may not be divided by large and substantially uniform WBC weighting, as is common in some SENSE reconstruction implementations and the like.

  In the present embodiment, the magnitude prescan expanded image “MPS” that is a composite image has substantially the same T1 and T2 weighting as the EPI main scan.

  EPI reconstruction is performed at least twice. The first pass reconstruction is performed without explicitly correcting for displacement or spatial distortion errors.

  In the simplest case, EPI main data is reconstructed twice in succession.

  Here, it is assumed that the number of reconstructions is two. In the second pass, another complete reconstruction is performed, but this time the spatial distortion correction is applied to the EPI ghost correction map and the SENSE or SPEDER coil sensitivity map.

  It is also possible that more iterations can make it somewhat more accurate. As the number of iterations increases (eg, 3 or more), the magnitude domain scan image “ME” should have less Nyquist ghosts and therefore better overall match the prescan image “MPS”. Increasing the number of iterations may reduce the need to recalculate the prescan image. This is because the ghost of the pre-scan image is removed from the image with few phase encoding blips and subsequent re-expansion in the image area.

  In the case of parallel imaging, the application to the main scan image is performed such that the original image associated with each coil or each channel region is folded in the phase encoding direction. For acquisition of the main scan image, it is beneficial to apply a stronger phase encoding gradient magnetic field with a factor that is a parallel imaging acceleration factor (or parallel imaging factor) “R”. The acceleration coefficient is generally in the range of 2 to 3, but considering the signal to noise ratio, RX coil geometry, etc., a fractional acceleration coefficient such as 1.6 or 4.0 A large acceleration factor may be used.

  Note that the parallel imaging factor necessitates the use of a larger phase encode gradient slope blip and an effective phase encode FOV that is smaller than the nominal phase encode FOV associated with the user defined effective final FOV. If this is the case, this factor is different from the damping factor associated with the EPI pre-scan gradient blip. For example, assume that the pre-scan gradient magnetic field blip is ½ the nominal value. It is also assumed that the parallel imaging coefficient is R = 3.0. In this case, the sum of the phase encode blip differences is a total coefficient (“AG”) of 6.0. This means that the geometric distortion of the EPI main scan (parallel imaging) is 6.0 times that of the pre-scan image.

  Note that individual information associated with individual coil sensitivities is required to perform parallel imaging development.

  A parallel imaging reconstruction similar to that of Non-Patent Document 1 is shown, but extended to use “pseudocoils”. The implementation of “pseudocoils” is described in Non-Patent Document 5.

  “N” and “M” are the sizes of the desired final image in units of pixels. “R” is a parallel imaging acceleration factor applied along the phase encoding direction in 2D parallel imaging.

  There can be various weightings or normalizations associated with the SENSE reconstruction. It is very beneficial that the spatial weighting in the pre-scan image “MPS” has the same weight as the magnitude domain scan image “ME”.

  Preferably, the matrix inversion is normalized to produce a pseudo-inversion matrix MINV, and the encoding matrix (output “MAT”) is numerically close to rank R1 <R (ie, the “Rth” singularity of MAT Avoid excessive noise gain in situations where the value is much smaller than the maximum singular value.

  The process loops for each pixel in the FOV reduced version of the image, ie in the folded image as opposed to the unfolded image.

  A spatial distortion (DPE) between the EPI pre-scan and the EPI main scan (parallel imaging) is calculated. The distortion amount can be used to directly distort the ghost correction map, and as a result, the ghost correction map has the same distortion as the EPI main scan (parallel imaging). Then, by changing the magnification of this distortion map, it is possible to estimate how much the coil sensitivity map is distorted as a result with respect to the EPI main scan (parallel imaging) image. Next, the coil sensitivity map is distorted so that the coil sensitivity map also has a geometric distortion that matches the distortion of the EPI main scan (parallel image) using the scaled distortion (DPE2).

  The number of repetitions may be two or more.

  For both the magnitude pre-scan stretched image (MPS) and the main scan image (ME), the signal is integrated for each column of images along the phase encoding direction.

  It may be beneficial to optionally normalize the two line integrals. For example, if the pre-scan image is the first acquisition, and if the EPI main scan following the pre-scan is slightly saturated because TR does not allow full recovery of the MZ equilibrium magnetization, The integral may be somewhat smaller than the total integral of the MPS image signal. In such a case, perhaps in a particular column, the ratio of the total line integral across the PE FOV may be determined, which is the value “LP” for the prescan full line integral and the EPI main scan total line integral value “LE”. " The line integrals may be normalized with respect to each other. For example, each pixel of the line integral obtained from the MPS column may be divided by the final denormalized value of LP, and each pixel of the line integral obtained from the ME column may be divided into the final denormalized LE. You may divide | segment by a value. Other than just normalizing the ME line integral for each column by the (LM / LE) coefficient, or just normalizing the MPS line integral for each column by the (LE / LM) coefficient, etc. These modifications are obviously possible and are conceivable in the embodiment.

  For each pixel position “Y” along each column of the line integration image LE, a value of line integration is obtained (denoted as a value “VEY”). Then, MPS line integration may be performed by changing the direction along the same column. If the values of MPS and ME are magnitude pixels (non-negative real values), and if each line integral is normalized and the value is in the range of zero to 1.0, then each line integral is considered as a monotonic function For example, if interpolation is performed along the line integral of the MPS column, a matching value can be reliably obtained.

  For a predetermined “Y” position in LE, an amplitude “VEY” is determined. Then, a search for the value “VEY” is performed along the line integral of MPS. Considering the possibility of interpolation of increasing values in the line integral of the MPS column, a new “YP” value, ie the corresponding y position, is obtained. The corresponding y position further divides the image so that all the signals of “VEY” are “below” of the dividing point “YP” and all the signals (1-VEY) are “above” of the dividing point. To do. This point “YP” is an appropriate estimate of the position of the corresponding original point “Y” if LE and LM change only due to the distortion difference. Therefore, the displacement distance “DPE” = (Y−YP), which is the distortion difference between the two images, can be generated from the point YP in the pre-scan or from the point Y in the EPI main scan.

  Other distortion estimates are also considered valid and are considered embodiments. For example, a 2D variable distortion software function may be used to suppress distortion that becomes a displacement field along only the phase encoding direction (not the PE direction or the readout RO direction), and it is locally smooth and mutual information. A distortion map that raises the metric to an appropriate limit may be determined. However, the above-described line integration technique allows the YP search to proceed starting from the position of the immediately preceding YP, and can process all DPE values along one column in an incremental manner, so that the calculation process is fast and highly efficient. is there. Thus, both Y and YP “walk” in a single direction across each line integral image. At that time, “Y” advances the pixels evenly one by one, and “YP” is expected to move approximately one pixel at a time, but in an uneven manner.

  It can be said that the displacement map “DPE” is substantially required to make “MPS” coincide with “ME”. However, since “MPS” is the same or similar physical measurement value as the distortion correction map “PHS”, the PHS is applicable to the same or similar distortion operation as “MPS”. The distortion map “DPE” can be applied to distort “PHS” along the phase encoding direction (along the columns), resulting in nominally the same distortion as “ME” and “CE”.

  An improved PHS can also be used.

  Note that distorting “MPS” is more preferable than trying to distort the folded images of “even_main” and “odd_main”. This is because each pixel in “even_main” and “odd_main” actually corresponds to a superposition of various folded pixels, and the displacement distance “DPE” is appropriate for one pixel in the folded set. This is because it is generally not appropriate for other folded pixel positions.

  The distortion in MPS (and PHS) is “AG” times the distortion in ME (and CE). However, the distortion in the coil sensitivity map is usually much smaller and can be considered zero for comparison. Therefore, if the strain DPE (Y) has a specific value, it may be considered that DPE (Y) = (AG * (DPE0) -1 * (DPE0)). Here, DPE0 is the absolute strain in ME. AG is well known as a simple constant. Thus, the above may be solved for DPE0. Specifically, DPE = (AG-1) * DPE0. Therefore, the distortion of each “map” from the ME is DPE2 = DPE * (− 1 / (AG−1)).

  DPE2 is applied to each sensitivity map.

  A new pseudo coil sensitivity map is created by combining the improved versions of PHS2 and map2. Here, “map2” and “PHS2” corrected to have substantially the same distortion as “CE” and “ME” are used. Therefore, “map2” and “PHS2” should both be derived from the same (distorted) positions as “CE” and “ME”, so the spatial between various maps and EPI main scan data There should be no error due to misregistration and have accurate values to perform reconstruction.

  The embodiment described above with reference to FIGS. 1 to 7 makes it possible to remove Nyquist ghosts with EPI, with or without parallel imaging, tolerant to off-resonance, or misalignment between various maps and EPI main scans. It is possible to obtain resistance against the ease of occurrence. EPI prescan can be used for both Nyquist ghost removal and geometric distortion reduction.

  Compared to the prior art that relies on the correction applied to the main scan image, the embodiments disclosed herein apply the correction to the lower resolution reconstruction map, thus increasing the tolerance to the spatial jitter that occurs in the correction. Technology is obtained. For example, if the quality of the spatial correction is reduced, the effect of ghost removal is reduced, but since the ghost is already at a low level, this is compared with the error caused by directly developing the main scan image without considering various distortions. This is a secondary effect.

  In the embodiment, the ghost is reduced, but unlike the conventional EPI space correction method, there is no mixing of the main scan information content between two or more shots of data collection.

  Furthermore, since the EPI pre-scan has already been executed for the normal EPI scan, the total scan time does not become long even though the correction more than the conventional EPI is executed. The EPI pre-scan is generally performed for 1 to 2 shots at the start of the EPI scan and is not presented to the user. Plural uses such as reduction of ghost, achievement of balanced magnetization of signal, securing of T1 balanced magnetization (T1 / TR contrast normalization), determination of optimum reception gain, and the like are possible. By using the same EPI pre-scan in determining the distortion map, the embodiment achieves the performance described above without increasing the scan time.

  Furthermore, in a time series, for example, perfusion application of FMRI, ASL, or DCE, there is no fluctuation for each shot. True time variation is not “filtered”, “blurred” or “adjusted”. One distortion map may be applied to a plurality of time-series EPI main scans.

  The Nyquist correction is a high-order spatial correction, and thus it is possible to correct an error accompanied by a spatial variation such as a local eddy current, a mechanical vibration, or a frequency dependency varying in the coil array.

  The final EPI image can then be subjected to various post-processing functions such as partial k-space homodyne filtering correction, high frequency image homogenization correction, tilt distortion correction and the like. A final diagnostic image is obtained from the final EPI image or the post-processed final EPI image.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and the equivalents thereof.

20 Magnetic Resonance Imaging (MRI) Equipment 22 MRI System Control Unit 30 MRI Sequence Control Unit 42 MRI Data Processing Unit

Claims (8)

A sequence control unit that executes a sequence including a pre-scan and a main scan on the subject;
Generating a pre-scan image and a correction map from the data collected by the pre-scan, and generating a main scan image from the data collected by the main scan;
Based on the geometric distortion difference between the main scan image and the pre-scan image, the correction map is modified to have substantially the same distortion as the geometric distortion in the main scan image. Correction part,
A magnetic resonance imaging apparatus comprising: a correction unit that corrects the main scan image using a corrected correction map. The sequence control unit further performs a pre-scan for the coil sensitivity map,
The generating unit further generates a coil sensitivity map from data collected by pre-scanning for the coil sensitivity map,
The correction unit further corrects the coil sensitivity map to have substantially the same distortion as the geometric distortion in the main scan image based on the distortion difference;
The correction unit further corrects the main scan image using the corrected coil sensitivity map.
The magnetic resonance imaging apparatus according to claim 1. The sequence control unit further executes a pre-scan for parallel imaging,
The generation unit further generates a folding map for development from the data collected by the prescan for parallel imaging,
The correction unit further corrects the folding map to have substantially the same distortion as the geometric distortion in the main scan image based on the distortion difference;
The correction unit further corrects the main scan image by further using the corrected folding map,
The magnetic resonance imaging apparatus according to claim 1 or 2. The correction unit compares the main scan image with the pre-scan image and corrects the correction map using a first distortion map generated by obtaining a distortion vector amount at each position.
The magnetic resonance imaging apparatus according to claim 1, 2 or 3. The correction unit compares the main scan image with the pre-scan image, generates a first distortion map by obtaining a distortion vector amount at each position, and further generates the coil sensitivity map for the main scan image. The second distortion map is generated by changing the magnification of the first distortion map according to the distortion of the first distortion map, the correction map is corrected using the first distortion map, and the second distortion map is To correct the coil sensitivity map,
The magnetic resonance imaging apparatus according to claim 2. The correction unit compares the main scan image with the pre-scan image, generates a first distortion map by obtaining a distortion vector amount at each position, and further generates a first distortion map with respect to the main scan image. A second distortion map is generated by changing the magnification of the first distortion map according to the distortion, the correction map is corrected using the first distortion map, and the second distortion map is used. To correct the folding map,
The magnetic resonance imaging apparatus according to claim 3. The correction unit identifies a corresponding point between the main scan image and the pre-scan image using a line integration method, and corrects based on the distortion difference determined between the identified points. To
The magnetic resonance imaging apparatus according to claim 1, 2 or 3. The sequence control unit executes the pre-scan with a smaller blip of the phase encoding gradient magnetic field than the main scan,
The correction unit corrects based on the distortion difference determined after expanding the pre-scan image so as to have the same FOV (Field Of View) as the main-scan image,
The magnetic resonance imaging apparatus according to claim 1.